Videos may be analyzed based on their semantic content to determine an activity being performed during video capture. Analyzing video content may involve techniques that are computationally expensive.